Presentation is loading. Please wait.

Presentation is loading. Please wait.

1 Chapter 8 Inference and Resolution for Problem Solving.

Similar presentations


Presentation on theme: "1 Chapter 8 Inference and Resolution for Problem Solving."— Presentation transcript:

1 1 Chapter 8 Inference and Resolution for Problem Solving

2 2 Chapter 8 Contents (1) l Normal Forms l Converting to CNF l Clauses l The Resolution Rule l Resolution Refutation l Proof by Refutation l Example: Graph Coloring

3 3 Chapter 8 Contents (2) l Normal Forms in Predicate Calculus l Skolemization l Unification l The Unification algorithm l The Resolution Algorithm l Horn Clauses in PROLOG

4 4 Normal Forms l A wff is in Conjunctive Normal Form (CNF) if it is a conjunction of disjunctions: n A 1 Λ A 2 Λ A 3 Λ … Λ A n n where each clause, A i, is of the form: n B 1 v B 2 v B 3 v … v B n n The B’s are literals. l E.g.: A Λ (B v C) Λ (¬A v ¬B v ¬C v D) l Similarly, a wff is in Disjunctive Normal Form (DNF) if it is a disjunction of conjunctions. l E.g.: A v (B Λ C) v (¬A Λ ¬B Λ ¬C Λ D)

5 5 Converting to CNF l Any wff can be converted to CNF by using the following equivalences: (1)A ↔ B  (A → B) Λ (B → A) (2)A → B  ¬ A v B (3) ¬ (A Λ B)  ¬ A v ¬ B (4) ¬ (A v B)  ¬ A Λ ¬ B (5) ¬¬ A  A (6)A v (B Λ C)  (A v B) Λ (A v C) l Importantly, this can be converted into an algorithm – this will be useful when when we come to automating resolution.

6 6 Clauses l Having converted a wff to CNF, it is usual to write it as a set of clauses. l E.g.: (A → B) → C l In CNF is: (A V C) Λ (¬B V C) l In clause form, we write: {(A, C), (¬B, C)}

7 7 The Resolution Rule l The resolution rule is written as follows: A v B¬ B v C A v C l This tells us that if we have two clauses that have a literal and its negation, we can combine them by removing that literal. l E.g.: if we have {(A, C), (¬A, D)} l We would apply resolution to get {C, D}

8 8 Resolution Refutation l Let us resolve: {(¬A, B), (¬A, ¬B, C), A, ¬C} l We begin by resolving the first clause with the second clause, thus eliminating B and ¬B: {(¬A, C), A, ¬ C} {C, ¬C} l Now we can resolve both remaining literals, which gives falsum:  l If we reach falsum, we have proved that our initial set of clauses were inconsistent. l This is written: {( ¬ A, B), ( ¬ A, ¬ B, C), A, ¬ C} ╞ 

9 9 Proof by Refutation l If we want to prove that a logical argument is valid, we negate its conclusion, convert it to clause form, and then try to derive falsum using resolution. l If we derive falsum, then our clauses were inconsistent, meaning the original argument was valid, since we negated its conclusion.

10 10 Proof by Refutation - Example l Our argument is: (A Λ ¬ B) → C A Λ ¬ B  C l Negate the conclusion and convert to clauses: {( ¬ A, B, C), A, ¬ B, ¬ C} l Now resolve: {(B, C), ¬ B, ¬ C} {C, ¬ C}  l We have reached falsum, so our original argument was valid.

11 11 Refutation Proofs in Tree Form l It is often sensible to represent a refutation proof in tree form: l In this case, the proof has failed, as we are left with E instead of falsum.

12 12 Example: Graph Coloring l Resolution refutation can be used to determine if a solution exists for a particular combinatorial problem. l For example, for graph coloring, we represent the assignment of colors to the nodes and the constraints regarding edges as propositions, and attempt to prove that the complete set of clauses is consistent. l This does not tell us how to color the graph, simply that it is possible.

13 13 Example: Graph Coloring (2) l Available colors are r (red), g (green) and b (blue). A r means that node A has been coloured red. l Each node must have exactly one color: A r v A g v A b ¬A r v ¬A g (  A r → ¬A g ) ¬A g v ¬A b ¬A b v ¬A r l If (A,B) is an edge, then: ¬A r v ¬B r ¬A b v ¬B b ¬A g v ¬B g l Now we construct the complete set of clauses for our graph, and try to see if they are consistent, using resolution.

14 14 Normal Forms in Predicate Calculus l A FOPC expression can be put in prenex normal form by converting it to CNF, with the quantifiers moved to the front. l For example, the wff: (  x)(A(x) → B(x)) → (  y)(A(y) Λ B(y)) l Converts to prenex normal form as: (  x)(  y)(((A(x) v A(y)) Λ (¬B(x) v A(y)) Λ (A(x) v B(y)) Λ (¬B(x) v B(y))))

15 15 Conversion to Prenex detail l (  x)(A(x)→B(x))→(  y)(A(y)ΛB(y)) l ¬ (  x)(A(x)→B(x)) v (  y)(A(y)ΛB(y)) l ¬ (  x)( ¬ A(x) v B(x)) v (  y)(A(y)ΛB(y)) l (  x)¬ ( ¬ A(x) v B(x)) v (  y)(A(y)ΛB(y)) l (  x)(¬¬ A(x)Λ ¬ B(x)) v (  y)(A(y)ΛB(y)) l (  x)( A(x)Λ ¬ B(x)) v (  y)(A(y)ΛB(y)) l (  x) (  y) ( A(x)Λ ¬ B(x)) v (A(y)ΛB(y))) l (  x)(  y)(((A(x)vA(y))Λ(¬B(x)vA(y))Λ(A(x) vB(y))Λ(¬B(x)v B(y))))

16 16 Skolemization l Before resolution can be applied,  must be removed, using skolemization. l Variables that are existentially quantified are replaced by a constant:  (x) P(x) l is converted to: P(c) l c must not already exist in the expression.

17 17 Skolem functions l If the existentially quantified variable is within the scope of a universally quantified variable, it must be replaced by a skolem function, which is a function of the universally quantified variable. l Sounds complicated, but is actually simple: (  x)(  y)(P(x,y)) l Is Skolemized to give: (  x)(P(x,f(x)) l After skolemization,  is dropped, and the expression converted to clauses in the usual manner.

18 18 Unification (1) l To resolve clauses we often need to make substitutions. For example: {(P(w,x)), (¬P(y,z))} l To resolve, we need to substitute y for w, and z for x, giving: {(P(y,z)), (¬P(y,z))} l Now these resolve to give falsum.

19 19 Unification (2) l A substitution that enables us to resolve a set of clauses is called a unifier. l We write the unifier we used above as: {y/w, z/x} l A unifier (u) is a most general unifier (mgu) if any other unifier can be formed by the composition of u with some other unifier. l An algorithm can be generated which obtains a most general unifier for any set of clauses.

20 20 The Resolution Algorithm 1First, negate the conclusion and add it to the list of assumptions. 2Now convert the assumptions into Prenex Normal Form 3Next, skolemize the resulting expression 4Now convert the expression into a set of clauses 5Now resolve the clauses using suitable unifiers. l This algorithm means we can write programs that automatically prove theorems using resolution.

21 21 Horn Clauses in PROLOG (1) l PROLOG uses resolution. l A Horn clause has at most one positive literal: A v ¬B v ¬C v ¬D v ¬E … l This can also be written as an implication: B Λ C Λ D Λ E → A l In PROLOG, this is written: A :- B, C, D, E

22 22 Horn Clauses in PROLOG (2) l Horn clauses can express rules: A :- B, C, D l Or facts: A :- l Or goals: :- B, C, D, E l If a set of clauses is valid, PROLOG will definitely prove it using resolution and depth first search.

23 23 Skolemization Example (  x)([a(X) Λb(X)]→[c(X,I)Λ(  X)((  Z)[c(X,Z)] →d(X,X))])v(  X)(e(X)) (  x)([a(X) Λb(X)]→[c(X,I)Λ(  Y)((  Z)[c(Y,Z)] →d(X,Y))])v(  X)(e(X)) (  x)( ¬ [a(X) Λb(X)]v[c(X,I)Λ(  Y)((  Z)[c(Y,Z)] →d(X,Y))])v(  X)(e(X)) (  x)( ¬ [a(X) Λb(X)]v[c(X,I)Λ(  Y)((  Z)[ ¬ c(Y,Z)] vd(X,Y))])v(  X)(e(X)) (  x)([ ¬ a(X) v¬b(X)]v[c(X,I)Λ(  Y)((  Z)[ ¬ c(Y,Z)] vd(X,Y))])v(  X)(e(X)) (  x)([ ¬ a(X) v¬b(X)]v[c(X,I)Λ(  Y)((  Z)[ ¬ c(Y,Z)] vd(X,Y))])v(  W)(e(W)) (  x) (  Y)(  Z) (  W)([ ¬ a(X) v¬b(X)]v[c(X,I)Λ(¬ c(Y,Z) vd(X,Y))]v e(W)) (  x) (  W)([ ¬ a(X) v¬b(X)]v[c(X,I)Λ(¬ c(f(X),g(X)) vd(X,f(X)))])v e(W)) [ ¬ a(X) v¬b(X)]v[c(X,I)Λ(¬ c(f(X),g(X)) vd(X,f(X)))])v e(W) [ ¬ a(X) v¬b(X)vc(X,I)v e(W)] Λ [ ¬ a(X) v¬b(X)v¬ c(f(X),g(X)) vd(X,f(X))v e(W) [ ¬ a(X) v¬b(X)vc(X,I)v e(W)] Λ [ ¬ a(U) v¬b(U)v¬ c(f(U),g(U)) vd(U,f(U))v e(V)


Download ppt "1 Chapter 8 Inference and Resolution for Problem Solving."

Similar presentations


Ads by Google